STEEL enables high-resolution delineation of spatiotemporal transcriptomic data

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Spatial transcriptomics offers unprecedented insights into tissue architecture and cellular heterogeneity.
• Existing methods face challenges in accurately identifying distinct tissue domains and integrating diverse datasets.

Purpose of the Study:

• To introduce Spatial Transcriptome based cEll typE cLustering (STEEL), a novel unsupervised algorithm for spatial transcriptomic data analysis.
• To demonstrate STEEL's capability in high-resolution domain identification and multi-slide integration.

Main Methods:

• STEEL employs unsupervised manifold learning and graph-based clustering.
• It groups beads based on gene expression similarity and spatial proximity.

Main Results:

• STEEL accurately characterized fine structures in mouse brain datasets (10X Visium).
• It successfully integrated multiple tissue slides for comprehensive analysis.
• STEEL outperformed existing methods on high-density, low-efficiency Slide-seq datasets.
• It delineated developmental landscapes and cell differentiation dynamics in early mouse embryos.

Conclusions:

• STEEL provides a robust and scalable approach for spatial transcriptomic domain identification.
• The algorithm facilitates the reconstruction of detailed gene expression atlases.
• STEEL is poised to significantly impact the analysis of complex biological tissues and developmental processes.